Avoid unnecessary capitalization. i.e., "Direct Shebang Execution" -> "Direct shebang execution"
See https://cbea.ms/git-commit/ carefully

When writing Git commit messages, avoid formatting them like this:

Implemented a recursive shebang parsing loop
in the standalone loader's entry point.
It reads shebang interpreter lines
(including optional arguments like #!/bin/sh -x),
updates elf_path to point to the interpreter,
prepends them to guest_argv, and resolves the
host path recursively up to a depth of 5 levels.

At first glance, this reads more like a haiku than a commit message. Since Git conventionally allows up to 72 characters per line in the message body, wrap the text naturally while making good use of the available width. For example:

Implemented a recursive shebang parsing loop in the standalone loader's
entry point. It reads shebang interpreter lines, including optional
arguments like #!/bin/sh -x, updates elf_path to the interpreter, and
prepends them to guest_argv while resolving the host path recursively
up to a maximum recursion depth of five interpreter levels.

The goal is not to make every line exactly 72 characters long, but to avoid unnecessarily short lines. Wrap at sensible boundaries so that the text remains easy to read while making efficient use of the available line width.

I have implemented all the requested updates from the new code review comments:

1. Nested Shebang Resolution in sys_execve: Refactored the execve path in exec.c into a bounded recursive loop (resolving up to 5 shebang layers recursively). Bounded depth exhaustion returns -LINUX_ELOOP. Temporary materialized paths are unlinked at each iteration so they don't leak.
2. Safe argc check: Changed new_argc verification in exec.c to a safer size-based condition (argc > MAX_ARGS - prefix + 1) that checks the boundary before addition, preventing potential overflow.
3. Naming & Documentation:
   • Renamed elf_parse_shebang to elf_read_shebang in all declarations, headers, and invocations to clarify it performs both probing (detection) and parsing.
   • Dropped the redundant description comment inside elf.c.
4. Overlong/Unterminated lines: Used strpbrk(buf + 2, "\r\n") in elf.c to locate line endings. If no line ending is found and the buffer is completely filled, it returns -ENOEXEC.
5. Host-Side Unit Tests: Added a host-side unit test file test-shebang-host.c that exercises:
   • CRLF, LF, and CR dialects.
   • Scripts ending without trailing newlines (EOF).
   • Empty/blank shebang interpreters.
   • Non-shebang scripts.
   • Overlong/unterminated shebang lines.
   • The test was successfully linked and runs automatically with make check.

All changes compile cleanly and pass the native unit tests.

All changes compile cleanly and pass the native unit tests. You can review the details in walkthrough.md.

We are humans. Don't copy agent-specific reply here.

Do you have any tool to help format the git commit message?
I break line manually so some line longer than others and not well formatted

Thank @doanbaotrung for contributing!